atcomsystems.ca/forum Forums General General Telephone & Systems Information CAT5 > Fiber > CAT5 ...why bother???

Hi folks,
I have, what I think is, a fundamental physics question about communication signal behavior.

My ISP supplies a 10Mb service to my business. In my server room, my connectivity to the LAN is supplied via a CAT5 (I don't know if it's a CAT5 or CAT5e cable) cable that originates from my telco closet (which is about 25 yards away). The CAT5 cable plugs into a SonicWall appliance which feeds a Cisco 4506. The Cisco supplies connectivity to one subnet via CAT5 cables and connects to another patch panel via fiber cables, which in turn supplies connectivity to another subnet via CAT5 cables.

So here's my fundamental question - no matter what type of signal, whether it's electrical or wireless, once a signal slows down...it will never speed up, right? In other words, if you have a mixing of Gigabit, 100Mb and 10Mb devices in your network, once a signal hits a 10Mb device and then, let's say, a Gigabit device...the signal isn't going to slow to 10Mb and then accelerate to Gigabit speeds, etc. just because it CAN, right? Once it slows down to a particular speed, it will not accelerate.

Sooo...if all of this is correct, why intermix fiber and CAT5 cables on a network? If you can't afford to go ALL fiber, why bother? It seems like you'd be spending $$$ for a performance that you're not taking advantage of at all.

Comments?

Thank you,
Ed

You are correct. The slowest medium controls the end-to-end speed of the whole system.

Optical fiber cables are used for reasons other than obtaining higher speeds. They are not susceptible to EMF crosstalk, and they can transport signals further distances than the length restrictions inherent in copper media.

Thank you Arthur!

Point of clarification: if you have a LAN with a mixture of fiber, 100Mb and 10Mb...the whole network wouldn't run at 10Mb, right?...but the signal would slow down to 10Mb when it hits that segment of the LAN from that point forward only, wouldn't it???

Ed

The switch ports would negotiate with the device and run at a speed agreed to by both devices. So if your NIC for example is an old 10baset NIC then the computer to switch connection would only run at 10 meg. However all of the other ports may be gigabit ports and connected to gigabit NICs so all of those would run at 1000 meg.

Fiber to the desktop is usually out of the budget for most scenarios...unless you use OPM, like the government.

Arthur says "The slowest medium controls the end-to-end speed of the whole system" (which is what *I* thought was correct), but what tito1411 says seems to make sense - packets will usually autonegotiate with every device at every hop, so, effectively, the signal DOES accelerate to best possible speeds that are supported by a given device.

So if a LAN DID have some 10Mb legacy hubs, NICs, etc., intermixed with 100Mb and Gb equipment, the signal WILL, overall, speed up and slow down at each hop.

Yes? No?

Ed

Yes- to a point.

Let's say you have a 45 mb (T-3) service coming in to the site that connects to a router. The router feeds out to a number of 100mbs switches all feeding Cat 5e/6 stations. These stations can communicate with each other at 100mbs. When they go out to the WAN they go out at 45 mbs.

Then you've got another feed to a hub that's supporting Cat 3 cables. These cables communicate with each other at no more then 10mbs. When they connect to the stations at the 100mbs switches they communicate at 10 mbs. When they go out to the WAN they go out at 10 mbs.

Does this make sense?


Sam

It sounds like you may be confusing the medium with the Ethernet technology. You can run 10Mb, 100Mb, 1000Mb and 10Gb on either copper or fiber, so long as you have the proper components for each. Fiber has higher capacity, but as Arthur mentions, it may be used for other reasons entirely, such as immunity to EMF.

Back to your question: You can look at it another way. Your faster links will introduce less latency than your slow links. If your switch is receiving packets on port 1 from a 100Base-T link and forwarding out port 2 on a 1000Base-T link, the packets will not be slowed down as much leaving the switch as they would if you had another 100Base-T link. Of course the switch cannot forward packets that it hasn't received yet from that slower link. So the packets in this example are going to travel at 100Mb, and then 1Gb, but there will be longer intervals between packets on the 1Gb side as the switch waits for packets from the 100Mb side, so the effective speed is never faster than 100Mb.

What if you have devices on ports 3 to 24 that also want to talk to the device on port 2? Now it makes a lot of sense to have a bigger pipe going to that device to reduce the bottleneck.

Hopefully that makes sense...it's way past my bedtime.

This brings up a question.
lets go to the extreme.

Have 1 48 port gigabit switch talking to a port running 10 mb.

so in theory each port on the gigabit switch is sending 100 times the amount of the information than the 10 mb will be able to pass at a given moment.

Routers are able to Store the Surplus of packets until there is a lull in the traffic, but what about switches? To a point (usually far lower than a router).
So when this buffer is maxed and there isn;t any lull what will happen to the rest of the devices on that switch? Will everything slow down? or will most of the packets just get discarded (which would also give the appearance of a slowdown). I think it would just be the latter since as I understand it, once the negotiation happens, ports won't dynamically re negotiate to another speed.

Sam:

Yes, I *think* I see what you mean - basically, you're saying that as packets travel through any network, they WILL travel as quick as they can from hop to hop, only being restricted by either the medium that they're traveling on (10BaseT, 100BaseT, Fiber, etc.) OR the limitations of the equipment itself.

Is that about right?

Ed